During the sixth week, we started looking into incorporating timing into the agents' algorithms. We got some new code from IBM with a timing strategy implemented for the ZIP traders. I added the algorithms that Amy and I had been working with, and modified them for timing as well. Below are the results of the experiments I ran. The revision option was implemented slightly differently in the new code, but the idea is the same. The low performance of the ZICTs was later explained and fixed.
Results with the expiration policy but NO revision option:
Algorithm | Avg. Efficiency | Trade Ratio |
1. AGHT | 0.998633 | 0.991333 |
2. GDS(mem=1) | 0.998433 | 0.958833 |
3. CLIFFT | 0.998267 | 0.992167 |
4. ZIPT | 0.997400 | 0.993500 |
5. ARGT | 0.996500 | 0.994167 |
6. GDL(mem=1) | 0.994667 | 0.954167 |
7. AGH | 0.9994067 | 0.984000 |
8. PVTT | 0.993967 | 0.977000 |
9. CLIFF | 0.990833 | 0.983000 |
10. ARG | 0.989833 | 0.983167 |
11. ZIP | 0.988933 | 0.983000 |
12. PVT | 0.973967 | 0.968333 |
13. GDT(mem=1) | 0.963233 | 0.950833 |
14. ZIC | 0.914900 | 0.864500 |
15. ZICT | 0.680067 | 0.553667 |
Results with the expiration policy AND the revision option:
Algorithm | Avg. Efficiency | Trade Ratio |
1. CLIFFT | 0.998833 | 0.995500 |
2. ARGT | 0.998567 | 0.991833 |
3. ZIPT | 0.998300 | 0.994333 |
4. AGHT | 0.998167 | 0.994167 |
5. PVTT | 0.996500 | 0.980833 |
6. AGH | 0.995933 | 0.988500 |
7. CLIFF | 0.994433 | 0.989833 |
8. ARG | 0.993800 | 0.994167 |
9. ZIP | 0.992867 | 0.991333 |
10. PVT | 0.989433 | 0.984833 |
11. GDL(mem=1) | 0.981100 | 0.943833 |
12. GDS(mem=1) | 0.967833 | 0.910333 |
13. GDT(mem=1) | 0.933367 | 0.907833 |
14. ZIC | 0.909700 | 0.847667 |
15. ZICT | 0.678200 | 0.526000 |